The concept of document root is simply unknown to cern_httpd.
You just map a part of URL space to physical file system, and
the URL hierarchy is in no way bound to the physical file system
hierarchy. You can omit physical directories from between, you
can map many "URL directories" to a physical directory, and in
a pathologic case what seems to be a subdirectory could actually
be a parent directory.
This is all theory, and nobody really uses the rules like that,
but its possible. In other words having a document root, a single
one or many of them, just doesn't necessarily make any sense.
-- Cheers, Ari --